Investors reassessed the sustainability of debt-fueled AI capital expenditure on June 23, triggering a one-day rout in the semiconductor sector that erased roughly $1.4 trillion in market capitalization—one of the largest single-day corrections for any technology segment on record.
The Philadelphia Semiconductor Index, the sector's primary benchmark, fell 7.9 percent, pulling it back from levels that had represented a gain of more than 50 percent over the previous year. Leading chipmakers posted steep losses: Marvell Technology dropped 17 percent, Broadcom fell 12.6 percent, and memory-chip leaders Micron Technology and SanDisk each lost approximately 13 percent. Nvidia, the most valuable company in the sector by market capitalization, surrendered $279 billion in value in the session—even as its percentage decline, at 4.1 percent, was comparatively modest relative to its peers.
The selloff was part of a broader technology retreat that pushed the Nasdaq Composite down 2.21 percent and the S&P 500 down 1.44 percent. Declining issues outnumbered advancers by roughly 1.3-to-1 on the New York Stock Exchange. The CBOE Volatility Index climbed to 19.52, its highest level in more than a week.
Several factors converged to produce the selloff. Stronger-than-expected U.S. employment data reinforced expectations that the Federal Reserve would keep interest rates elevated for longer than markets had anticipated, a shift that disproportionately pressures valuations of high-growth, long-duration technology companies. Rising Treasury yields compress the present value of distant earnings—a particularly acute problem for semiconductor firms whose most profitable revenue lies years in the future as data center deployments scale.
But the dominant narrative among portfolio managers was something more specific: a growing skepticism about whether the staggering amounts of capital being deployed into AI infrastructure can be absorbed by demand at the pace investors have assumed. The AI investment thesis has depended on the assumption that hyperscalers—cloud computing giants such as Amazon, Microsoft, and Google—will continue to absorb ever-larger volumes of high-performance chips at premium prices. When questions about that assumption surfaced, the response was swift and compressed.
"Some of the news lately about AI raises questions about all the spending that's being done and the capex and ramping of the capacity for semiconductors," said Thomas Martin, a senior portfolio manager at Globalt. "That's what markets are reacting to today."
The correction did not extend uniformly across all sectors. Six of the eleven S&P 500 sectors finished with gains, led by consumer staples, as investors rotated assets out of high-multiple technology names into more defensive holdings—a pattern consistent with valuation reset rather than a broader loss of confidence in the economic outlook. Corporate earnings remained solid, and the underlying demand trajectory for AI chips has not structurally changed.
Analysts at multiple banks, including Stifel, BofA Securities, and KeyBanc, maintained buy ratings on the most affected companies in the days following the selloff, with several raising price targets. Jensen Huang, Nvidia's chief executive, publicly described the correction as a buying opportunity for investors with a long-term horizon. That posture reflects a view consistent with previous cycles in the industry, in which sharp sentiment-driven declines in semiconductor stocks have been followed by robust recoveries once data-center demand proved durable.
Still, the speed and scale of the drawdown served as a reminder that markets price expectations as heavily as they price realities. The AI semiconductor sector had risen so far, so fast, that tolerances for doubt were slim. When a credible enough narrative for caution appeared, the resulting liquidation was both violent and brief. Whether it proves to be a corrective pause or the early sign of a broader revaluation of AI infrastructure spending will depend on earnings and guidance over the coming weeks.
Image source: v3b.fal.media